An integrated MEMS three-dimensional tactile sensor with large force range

Tao Mei, Wen J. Li, Ge Yu, Yong Chen, Lin Ni, Ming Ho ChanSensors and Actuators A Physical 2000 · DOI

TL;DR

This paper presents a novel, integrated MEMS-based three-dimensional tactile sensor designed for robotic applications. It addresses the need for robust, compact sensors capable of measuring both normal and tangential forces with a large force range, overcoming limitations of previous designs.

What Problem Does This Paper Address?

Robotic systems, particularly those operating in challenging environments like space, require sophisticated tactile sensing capabilities. Existing tactile sensors often fall short in several key areas: they may be too fragile for demanding applications, lack the ability to measure tangential (shear) forces crucial for grip control and preventing slippage, or are too bulky and heavy, increasing payload costs for space missions. The authors aim to bridge this gap by developing a sensor that is simultaneously robust, compact, sensitive to three-dimensional forces, and possesses a large force range, making it suitable for advanced robotic manipulation.

Methods

The core of the sensor is a MEMS (Micro-Electro-Mechanical Systems) structure fabricated using silicon bulk-micromachining. The design features an "E-shaped" membrane with integrated piezoresistors. Finite Element Analysis (FEA) was used to optimize the membrane structure and the placement of these piezoresistors to achieve independent and linear responses to forces applied in the X, Y, and Z directions.

The sensor is constructed from five layers: a shock-absorbing rubber surface, force-concentrating columns, the sensing array itself, a protection base, and a circuit board. The force-concentrating columns are designed to channel applied forces to the center of the sensing cells, where the piezoresistors are located. An "over-stop" design, implemented through a protection block, limits membrane deflection to prevent damage from excessive forces.

Crucially, the sensor integrates CMOS circuitry directly onto the chip for sensing, analogue switches, and data readout. This integration significantly reduces the overall sensor volume and the number of external connections required. The fabrication process involves standard CMOS processing for the circuits and bulk-micromachining for the MEMS structures. Neural networks are employed for post-processing to compensate for sensitivity variations caused by the rubber layer and to achieve accurate three-dimensional force measurements.

Datasets

Not discussed in the paper. The paper focuses on the design, fabrication, and calibration of a single sensor prototype.

Key Insights and Ideas

  • Integrated MEMS and CMOS for Robustness and Compactness: Combining MEMS for sensing and CMOS for electronics on a single chip is a key enabler for creating small, lightweight, and highly functional tactile sensors.
  • "E-shaped" Membrane for 3D Force Sensing: The asymmetric strain distribution induced by forces on the E-shaped membrane is ingeniously exploited to derive independent X, Y, and Z force components from the piezoresistors.
  • Force Concentrating Columns and Overload Protection: The design effectively channels forces to the sensing elements while incorporating a mechanical overload protection mechanism, enhancing sensor durability.
  • Neural Network Compensation for Material Effects: Using neural networks to compensate for the complex mechanical interactions and sensitivities introduced by the soft rubber layer is a practical approach to achieving high accuracy.

Key Results

The developed tactile sensor achieves a large force range of 50 N in the Z direction and ±10 N in the X and Y directions. The sensing cells exhibit independent and linear responses to the three force components. The sensor achieved approximately 2% FS (Full Scale) accuracy for total tactile force measurements within its specified force range after neural network training. The sensitivity was measured at about 13 mV/N in the Z direction and 2.3 mV/N in the X and Y directions. The overall sensor size is compact, measuring 20 × 50 × 7 mm³.

Limitations and Future Work

  • Mechanical Analysis of Multilayer Structure: The authors acknowledge that the mechanical analysis of the complex multilayer structure (rubber layer, E-shaped membrane, silicon glue) is an ongoing research issue.
  • Neural Network Complexity: While effective, the neural network approach requires training and introduces computational overhead. The paper mentions a back-propagation network with 97 input neurons, 32 hidden neurons, and 3 output neurons.
  • Future Integration: The paper states that the integration of this tactile sensor into a space-tested robotic system is underway, suggesting future work will focus on real-world deployment and performance validation in a robotic context.

Summary updated

BibTeX
@inproceedings{mei2000integrated,
  author = {Tao Mei and Wen J. Li and Ge Yu and Yong Chen and Lin Ni and Ming Ho Chan},
  title = {{An integrated MEMS three-dimensional tactile sensor with large force range}},
  booktitle = {Sensors and Actuators A Physical},
  year = {2000},
  doi = {10.1016/s0924-4247(99)00261-7},
}
Figure